60 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO.

1, MARCH 2009

Sensorless Slowdown Detection Method

for Single-Phase Induction Motors
Frédéric Ferreyre, René Goyet, Guy Clerc, Member, IEEE, and Thierry Bouscasse

Abstract—The single-phase capacitor-run induction motor is

generally used for fractional power in applications where the price
is an important criterion. In order to save costs, one tries to esti-
mate the speed on the basis of voltage and current measurements,
and thus avoid a speed sensor. The techniques used are very var-
ied. In the last few years, they have included numerous patents and
often the use of microcontrollers. First of all, this paper shows the
analytical characteristics of the induction motor with its capacitor.
It explains the different possible methods to assess the speed. Then,
the paper suggests a realization using one of these methods on a
motor of a rolling shutter. The setup tested implements a microcon-
troller that cuts off the supply as soon as an abnormal reduction in
speed is observed on a basis of a voltage measurement. This paper
details the implementation on the industrial application.
Index Terms—Capacitor motor, induction motor, microcon-
troller, sensorless speed estimation, single-phase, two-phase.
I. INTRODUCTION
APACITOR-RUN single-phase induction machines
C (SPIMs) are very widespread in the applications supplied
by the electric network at low cost: pumps, fans, lawn mow-
ers, home automation, etc. In this frame, it often appears to be
necessary to know the speed from electric measurements. The
objectives are as follows.
1) Command objectives like the starting capacitor command
[1] or vector controls [2].
2) Safety objectives like operating at excessive load, jam-
ming or blockage. For instance, on rolling shutters, the
arrival of the shutter at the buffer stop results in a drop in
motor speed.
The most precise method to know the speed of an induction
motor consists in constructing a speed observer in transient
state [3], [4]. This requires heavy calculation means as well as
measurements of the two voltages and currents.
For compactness and low-cost requirements, such observers
can be expensive. The command circuit must need few electric
sensors and small processing means. The solution is to research
into the parameters that vary with the speed.
For these low-cost applications, there are different methods
to estimate the speed. All are based on the behavior of SPIM
Manuscript received November 28, 2007; revised April 11, 2008. Current
version published February 19, 2009. Paper no. TEC-00464-2007.
F. Ferreyre is with the Université de Lyon, F-69622 Lyon, France, with Univer-
sité Lyon 1, F-69622 Villeurbanne, France, with CNRS, UMR 5005, Laboratoire
Ampère, F-69621 Villeurbanne, France, and also with Profalux, Thyez 74300,
France (e-mail: frederic.ferreyre@profalux.com).
R. Goyet and G. Clerc are with the Université de Lyon, F-69622 Lyon,
France, with Université Lyon 1, F-69622 Villeurbanne, France, and also with
CNRS, UMR 5005, Laboratoire Ampère, F-69621 Villeurbanne, France (e-mail:
rene.goyet@univ-lyon1.fr; guy.clerc@univ-lyon1.fr).
T. Bouscasse is with Profalux, Thyez 74300, France (e-mail: thierry.
bouscasse@profalux.com).
Digital Object Identifier 10.1109/TEC.2008.2008936
Fig. 1. Supply device of a capacitor-run single-phase induction motor.
in steady-state operation. Occasionally mentioned in patents,
these methods are not yet theoretically justified and methodi-
cally compared. This paper suggests such an analysis.
SPIM operates with an unbalanced power supply, so the
analysis is based on the forward and backward components.
The method is classically used with SPIM, as well in previous
books [5] or papers [6]–[9], as in more recent papers [10], [11].
The present study uses the terms of the impedance matrix
and leads to the drawing of various characteristics represen-
tative of speed in steady state. These characteristics allow to
define the criteria for choosing between the methods described
in the patents [12]–[16]. A strategy is suggested to detect the
slowing down of the motor with few mathematical operations.
This strategy is carried out with the early detection of the arrival
of a rolling shutter at the buffer stop.
II. MODELING AND CHARACTERISTICS
The two windings of the motor (subscript “1” and “2”) and
the capacitor C are often supplied from the mains with a circuit
comprising two switches Sw1 and Sw2 (see Fig. 1). When the
motor is symmetrical and if necessary, closing one or the other
of the two switches regulates the rotational direction.
The voltages and currents of the stator are recorded with the
indices 1 and 2, and their forward and backward components
with the exponents + and −.
The modeling considers that the two windings are identical
(two-phase symmetrical induction machine) and does not take
the iron losses into consideration. In these conditions, SPIM
may be represented by only two inductances, LS and N , and
two resistances, RS and RR . Equivalent dipoles, for forward
and backward components, are those of Fig. 2(a) and (b) [5].
The primitive dipoles with three inductances are used in
[6]–[11], two for leakages and one for magnetizing, and in paral-
lel with the magnetizing inductance, a third resistance represents

0885-8969/$25.00 © 2009 IEEE

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
FERREYRE et al.: SENSORLESS SLOWDOWN DETECTION METHOD 61

s = 1 − (ωR /ω) is a function of the pulsation ω of the stator

currents and the rotor speed ωR

V+ jLs ω (jN ω + (RR /s))

Z+ = = Rs + (3)
I+ (RR /s) + j (N + Ls ) ω

and
Fig. 2. Equivalent dipoles for forward. (a) Backward. (b) Components.
V− jLs ω [jN ω + (RR /(2 − s))]
Z− = −
= Rs + . (4)
the iron losses. Such accuracy is not necessary in low-cost speed I (RR /(2 − s)) + j (N + Ls ) ω
sensor applications.
The motor impedance matrix is
Generally, the most relevant parameter of the induction motor




is the ratio (RR /N ω) between the rotor resistance (RR ) and the V1 A B I1
total leakage impedance (N ω), with ω being the pulsation of = (5)
V2 −B A I2
the stator currents equal to the synchronism speed in the case of
a single pair of poles. with A = (Z + + Z − )/2 and B = j[(Z + − Z − )/2].
In well-balanced steady-state operation, (RR /N ω) deter- If the phase 2 is supplied by the mains, in parallel with the
mines the shape of the torque curve Γ(ωR ), ωR being the phase 1 connected in series with the capacitor C, this gives
rotor speed. For (RR /N ω) < 1, Γ(ωR ) features a maximum
V2 = V1 + ZL I1 (6)
for ωR < ω. On the contrary, when (RR /N ω) > 1, Γ(ωR ) de-
creases uniformly when ωR increases from 0 to ω. The motors where ZL = 1/jCω.
with a rated power of more than 1 kW, with high efficiency, are The preceding equations result in some “specific ratios” that
rather of the first type, with low RR . The motors with a rated facilitate characterizing the operational points
power of less than 1 kW for automatic control systems are often
I1 A−B
of the second type. = (7)
In unbalanced operation with a running capacitor, the value I2 A + B + ZL
of (RR /N ω) also influences the choice of the parameters to be V1 A2 + B 2 + B × ZL
= 2 (8)
used to estimate ωR as finely as possible. The study draws the V2 A + B 2 + A × ZL
main tendencies of the phenomenon, and justifies the different
V1 A2 + B 2 + B × ZL
methods to estimate ωR , adopted in industrial realizations [13]– Z1 = = (9)
[17]. The method also depends on the structure of the command I1 A−B
device. For example, the switches Sw1 and Sw2 that regulate V2 A2 + B 2 + A × ZL
the rotational direction in Fig. 1 can be placed at the bottom Z2 = = (10)
I2 A + B + ZL
on the neutral side as well (point O). This configuration influ-
I 1 2A + ZL
ences the choice of the method to estimate the rotor speed ωR . = = 2 (11)
V2 Z A + B 2 + A × ZL
A. Forward and Backward Components where I = I1 + I2 is the total current absorbed by the motor.
The mean power PR absorbed at the rotor includes Joule
Calculations are turned to express the voltages and currents
effect losses and mechanical power. PR is a function of the
of the two windings, 1 and 2, according to the rotor speed ωR .
currents I1 , I2 , I + , I − , and the voltages V
at the terminal of
Conversely, ωR can be estimated from voltages and/or currents
LS inductances (see Fig. 2).
measurements. The system (1, 2) is broken up into its forward
1) There are two V
voltages for the real motor:
and backward symmetrical components (+, −) with the com-
plex representation [5]. The norm of the numbers represents the V1
= V1 − RS I1 V2
= V2 − RS I2 .
amplitude of the figures and not their rms value. The equations
for the currents are 2) There are two V
voltages for the equivalent dipole of
Fig. 2:



+ 
I1 1 1 I
= (1) V
+ = V + − RS I + V
− = V − − RS I −
I2 −j j I−

+

 1
I 1 1 j I1 PR = Re[V1
(I1∗ ) + V2
(I2∗ )]
= (2) 2
− 2 1 −j
I I2 = Re[V
+ (I +∗ ) + V
− (I −∗ )]. (12)
and the same for the voltages. Re indicates “real part” and ∗ indicates “conjugate.”
The direct impedance Z + and the inverse impedance Z − can The torque is the difference between its forward and backward
be expressed depending on the two LS and N inductances and components
the two RS and RR resistances. The expressions are given by
the equivalent dipoles of Fig. 2(a) and 2(b) [5], in which the slip Re [V
+ (I + )∗ ] Re [V
− (I − )∗ ]
T = + (13)
ω −ω

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
62 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 1, MARCH 2009

1 +2 TABLE I
T = (|I | Re(Z + − Rs ) − |I − |2 Re(Z − − Rs )) R S , L S , N , AND R R AS A FUNCTION OF THE TEMPERATURE θ
ω
 2   
1  V +   +   V − 2  − 
T = Re Z − Rs −  −  Re Z − Rs .
ω  Z+  Z

In other words, by introducing the rms value V2rm s of the ac

power voltage V2 (t)
 2
V2rm s 2  (V1 /V2 ) + j   
T =  +  Re Z + − RS
2ω Z
  
 (V1 /V2 ) − j 2  − 
−   Re Z − RS
 . (14)
Z−

This expression for T slightly differs from [9] and [10], but it
can easily be shown that, with no stator leakage inductance and
no difference between the windings 1 and 2, the expressions are
similar.
Some remarks about the power transfer and the mathematical
transformations are as follows.
1) In (1) and (2)



1 1 1 1 j
P = P −1 =
−j j 2 1 −j
only separate the forward and backward components on
condition that I1 and I2 are expressed with the complex
representation.
2) The transformations P and P ∗ are such as


2 0
PT P ∗ = .
0 2
This gives V1 (I1∗ ) + V2 (I2∗ ) = 2(V + (I +∗ ) + V − (I −∗ ))
and a factor 1/2 in (12). Another transformation



1 1 1
P =√
2 −j j
may be used. It is such as



1 1 j 1 0
P
−1 = √ PT
P
∗ =
2 1 −j 0 1
and then the factors 2 and 1/2 disappear in complex power Fig. 3. (a) Torque and arguments of differing complex numbers characteristics,
expressions. ratios of voltages, currents or impedances, depending on the reduced speed x =
ω R /ω = 1 − s. (b) Amplitude of differing complex numbers, ratios of voltages,
currents or impedances, depending on the reduced speed x = ω R /ω = 1 − s.
B. Voltage and Current Curves for a Rolling Shutter Motor
Although the rolling shutter SPIM is geometrically and mag- nificance of the iron losses and shows the limit of an identifi-
netically balanced, its identification from four parameters is not cation from four parameters. We have, however, adopted this
obvious. First of all, a fifth parameter is theoretically necessary four-parameter model as it is practically analytical and makes it
to represent the iron losses. Then, the parameters, specifically easier to represent in a sufficiently accurate manner the various
the stator resistance, depend on the temperature. The identifi- operational regimes.
cation is nevertheless possible from experimental current and The analytical model is used to calculate the curves of Fig. 3.
voltage measurements. It is made in order to maximize the con- The four parameters are those of the first line of Table I, at
formity between measures and calculated data, from zero speed θ = 25 ◦ C, and the capacitor is C = 4 µF.
to synchronism at every temperature. This gives the results of Fig. 3(a) and 3(b) confirms the variations, depending on the
Table I. operating point, of several “specific complex numbers” (7)–
A dc measurement of the stator resistance gives a value (11), shown in the preceding paragraph. Fig. 3(a) gives their
250 Ω quite less than RS in Table I. This underlines the sig- arguments and Fig. 3(b) their amplitude. The relative speed

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
FERREYRE et al.: SENSORLESS SLOWDOWN DETECTION METHOD 63

the relative speed x varies. Fig. 5 gives the configuration of

the voltage triangles obtained for three motors, at two different
speeds: x = 0 (notation
) and x = 1 (notation without
). Fig. 5
also gives the three complex currents: I1 , I2 , and I = I1 + I2 .
The Fresnel diagram Fig. 5(b) corresponds to the rolling
shutter SPIM of this paper with θ = 25 ◦ C, C = 4 µF, LS 0 =
1.535 H, N0 = 0.072 H, RS 0 = 275 Ω, and RR 0 = 475 Ω.
The two other motors of Fig. 5(a) and (c) are close to this one.
As explained in Table II, one or two parameters are changed.
The motor of Fig. 5(a) has a lower inductance LS = 1.195 H
instead of LS 0 = 1.535 H.
The motor of Fig. 5(c) has lower stator and rotor resistances:
RS = 41 Ω instead of RS 0 = 275 Ω. The assumption is realistic
Fig. 4. Ideal conditions without backward component. as, for rolling shutter tubular motors, the size is reduced and RS
x = ωR /ω = 1 − s is shown on the X-axis. These analytical is usually high. Besides, it facilitates the running on locked rotor
results turn out to be very close to the measurements. Indeed, at nominal voltage.
even if the model is not exactly realistic, the parameters are The motors in Fig. 5(a) and (c) have a specific behavior. The
identified for their concordance. Table II highlights this behavior comparing it to the case of
The torque T (x) shows a maximum for x = 0.2 [see Fig 3(a)]. Fig. 5(b).
This comes from the unbalanced regime. On the contrary, with For the motor in Fig. 5(a), the points C, for x = 1, and C
, for
a balanced power supply, the ratio (RR /N ω) = 21 is fairly x = 0, are on the same straight line CA = C
A. The two angles
high to eliminate any maximum in the torque curve T (x) (as



mentioned in Section II). V2
V2
(CAO) = arg (C AO) = arg
With C = 4 µF, no speed meets the ideal condition that elim- VC VC

inates the backward component and leads to a global power
factor equal to 1. This ideal regime is designated by


 are equal (38◦ ), and their measurement does not make it possible
V2 −VC
α1 = arg(Z1 ) = arg(Z2 ) = arg = arg = 45◦ to estimate the relative speed x. In fact, during the passage from
VC V1 x = 1 to x = 0, the trajectory of point C deviates a little from



I1 V1 the straight line passing through A, but the measurement of x
α2 = arg = arg = 90◦ by (CAO) remains not very sensitive.
I2 V2

       In the same way for Fig. 5(c), the two angles for x = 1 and
V2  I1   V1   Z1 
α3 = arg = 0 k =   =   =   = 1
◦ x = 0, (AOC) = arg (V1 /V2 )and (AOC
), are equal (98◦ ), and
I1 + I2 I2 V2 Z2 the measurement of the relative speed x by (AOC) is not possi-
and corresponds to the Fresnel diagram in Fig. 4 without back- ble. The motor is “low” resistant. Instead, in this case, the angle
ward component. (AOC) can be useful for determining approximately when the
However, another capacitance C1 may give this ideal regime torque is maximal, especially in the management of capacitors
for a specific speed x1 and a specific torque value. The values of in capacitor-start induction motors [1].
C1 and x1 are C1 = 3.72 µF and x1 = 0.65. Given this operating The situation where the two angles (OCA), for x = 1, and
condition, Fig. 3 presents a quite different shape. (OC
A), for x = 0, are equal also exists. It is found, for example,
In Fig. 3(a), the group of points around B and C, inside the in the case of LS = 1.3 LS 0 , N = 4.17 N0 , RS = 0.098 RS 0 ,
circle, joins together. This unique point, the point A, and the and RR = 0.098 RR 0 . One finds OCA = OC
A = 36◦ . The Fres-
point D are aligned on the same vertical with the characteristics nel diagram is not shown in Fig. 5. In this case, the measurement
α1 = 45◦ , α2 = 90◦ , and α3 = 0◦ . of x by (OCA) is not possible.
In Fig. 3(b), the ordinate of A would be k = 1. In the case of the rolling shutter SPIM of this paper [see
The nominal speed of the SPIM of this paper is very close to Fig. 5(b)], the angle (CAO) changes from 47◦ for x = 1 to
these ideal operating conditions. 42◦ for x = 0. It does not, therefore, vary very much, and its
The curves Fig. 3 facilitate the estimate of the variations of measurement is not advisable for an estimation of x.
each value with the speed x. These variations are more often The differences of reaction between motors explain the large
than not monotonic, and each value can be used to assess the number of patents lodged on the subject. The discussion is not
relative speed x. Research into the most accurate measure is easy in as much as few details are given on the motors in ques-
made easier by plotting the Fresnel diagrams. This is done in tion. In spite of everything, we can suggest these simple expla-
the next paragraph. nations, for example, in the following two patents: Tecumseh
makes pumps and cold section refrigerating motors. They sug-
C. Fresnel Diagrams for Different Motors
gest measuring the relative x speed on the basis of the angle
In the complex plane, the three voltages V2 = OA, V1 = (CAO). The explanation is that RS is low enough for the mea-
OC, and VC = CA form a triangle OAC that is distorted when surement to be sensitive [12].

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
64 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 1, MARCH 2009

Fig. 5. Fresnel diagrams of voltages and currents with their trajectories for x ∈ [0; 1]. Scale: for currents “1” means 1 A, and for voltages “1” means 200 V. The
vectors at the synchronism speed (x = 1) are noted without asterisk (
) and those of null speed (x = 0) are noted with an asterisk (
).

TABLE II
RELEVANCE OF THE MEASUREMENTS DEPENDING ON THE TYPE OF MOTOR

Bubendorf makes “high RS ” rolling shutter motors, which
avoids the measurement of x by (CAO) and suggests (AOC)
[14].
A third patent, Plumer Company [13], gives all the possi-
ble solutions for the measurement of relative speed x with
the phase differences (arguments) of the complex numbers
(7)–(11), but it does not confirm how to choose one rather than
the other. The company insists rather on the signal process-
ing techniques to estimate the variations of the value measured
(sliding derivative).
European patents in [15] and [16] suggest to use the phase
between voltages and currents in order to prevent the blockage
of SPIM.
In conclusion, with SPIM, there are many relevant measure-
ments of phase or amplitude, of voltages or currents, to detect
the speed. Nevertheless, the speed estimation by the voltages is
generally preferable for reasons of cost.
III. SLOWING-DOWN CRITERIA
A. Method
The data processing of measurements in rolling shutter mo-
tion is complicated due to many reasons.
1) The motor parameters change with the temperature.
2) The running of the slats is very irregular.
3) There are many different kinds of motors and shutters.
Measures have to be compared with thresholds and filtered
in order to eliminate doubtful acquisition and noise. There are
certainly many suitable signal processing methods for such elim-
ination. The one presented in this paper is not so simple but has
the merit of being safe for a wide range of shutters and operating
conditions.
The previous study has shown that for rolling shutter
SPIM, the speed could be determined on notably the basis of
arg (V1 /V2 ) or |V1 /V2 |, values that are adapted for these values
decrease if the speed decreases [see Fig 5(b)]. Therefore, they
are used to detect an excessive decrease in speed and cut off the
supply of the motor. In this application, the decrease represents
the arrival of the rolling shutter on the buffer stop.
The same signal processing, applied to the two values, leads
each time to a stoppage criterion. The stoppage is effective if
the two criteria are checked simultaneously. In the case of the
rolling shutters, it has appeared that this “AND” function reduced
the risk of spurious stoppage.
The criterion used is explained later by showing indifferently
by “y” one or other of the two measurements. These measure-
ments are renewed every ∆t = 10 ms, the half-period of the
50 Hz supply. At the instant ti , the measurement yi can be
analyzed in several ways. For example, it is possible:
1) to calculate the absolute difference (yi − y0 ), between yi
and a fixed low value y0 , very much lower than the nominal
operational conditions. This technique is not definite as
there is a risk of the reaction being too late;

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
FERREYRE et al.: SENSORLESS SLOWDOWN DETECTION METHOD
Fig. 6. Input signal and output envelopes for the noise reduction stage-–(Env
max) and (Env min) surround y; y: arbitrary vertical scale.
2) to compare the relative difference between two consecu-
tive measurements and a threshold S; in other words, to
calculate
E (i) = [(yi − yi−1 ) − S] .
Stop the motor when E(i) > 0. This technique is not
suitable either as, whatever the value of S, even very low,
there is always an insignificant slope
yi − yi−1
pi =
∆t
repeated at each ti , so as to always have E(i) < 0. In this
way, it is possible to attain 0 speed without cutout. Even
more, with a threshold S too small, there is a risk that the
system might stop on the noise component.
3) To calculate several values
E (i, j) = [(yi − yi−j ) − S]
with j varying from 1 to N . Stop the motor at the first j = J
if E(i, J) > 0. The threshold S and the number of times N
where E(i, j) is tested at each value of i are predefined. N
and S depend on pi . With the rolling shutters, the situation
is complicated by the irregularities in speed caused by
the unwinding of the slats. The choices of N and S are
delicate; they can even require to consider functions N (i)
and S(i, j).
The third method has been chosen with fixed N = 18 and
variable S with j, which is referred to as S(j).
B. Implementation
The measurements, recorded every 10 ms, are very noisy.
Fig. 6 gives an example of a typical measure of arg (V1 /V2 ) or
|V1 /V2 |; both are similar, and the scale is arbitrary. The measure,
named y, illustrates the signal processing carried out. The latter
comprises three stages 1)–3).
1) Reduction in Noise: The noise reduction has been first
realized with filtering techniques, like sliding average over a
great number of values of y. It appears that the method involves
a long, unacceptable delay, in comparison with the desired reac-
Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
65
tion time of the system. Then, another method has been adopted.
It consists of replacing the noisy y signal by two other signals
without noise surrounding y, called EnvMax and EnvMin that
are the lower and the upper bound of y.
For that, at each yi , three values are calculated
yi + yi−1
sliding average: Syi = (16)
2
|yi − yi−1 |
difference: Dyi = (17)
2
thickness: Eyi = max (Dyi ) . (18)
These three values are a function of the evolution of y and
the background noise. They are used to calculate the envelopes,
EnvMax and EnvMin. The calculation is made on the basis of
the following logic.
1) If Syi > EnvMaxi , then
{EnvMaxi+1 = Syi and EnvMini+1 = Syi − Eyi }.
2) If Syi < EnvMini , then
{EnvMini+1 = Syi and EnvMaxi+1 = Syi + Eyi }.
Fig. 6 shows how EnvMini and EnvMaxi surround y.
The following study is carried out on the basis of EnvMini .
During a continuous decrease in y, EnvMini follows Syi , which
itself is very near to yi . Once initialized in the first measure-
ments, the thickness Eyi hardly changes at all in practice in
such a way that the system operates as a hysteresis follower
with two outputs, one being the limit inferior and the other
the limit superior of y. When the instantaneous variation of y
exceeds Eyi , EnvMin updates this variation with the smallest
possible delay (10 ms).
2) Buffering and Detection: The reduction in speed is stud-
ied from N successive measurements of y stored in the form
of
(15) αi EnvMini (19)
with αi = 1 if y decreases, i.e., if (Syi < EnvMini−1 ), and αi =
0 in the opposite case.
This arrangement means taking into account only the decreas-
ing measurements of yi . This has turned out to be beneficial in
the case of rolling shutters to reduce calculation time and elim-
inate the irregularities of speed.
After each EnvMini storage, N comparisons of αEnvMin to
N different thresholds S(j) are made.
The motor is stopped if the following condition is verified:
(αi−j EnvMini−j − EnvMini ) > Sj . (20)
The comparison is not made if (αi−j EnvMini−j = 0).
Fig. 7 indicates the process. At the top, the storage memory
is presented with only 13 of the 18 cells.
Later, the 13 corresponding thresholds are given that represent
the maximum differences of speed permitted. The differences
(αi−j EnvMini−j − EnvMini ) = (αi−j EnvMini−j − 1039) be-
tween the two are indicated each time they are calculated. The
first time the difference equals 12, the threshold 13, and move on
to the second calculation. At the fourth calculation, the differ-
ence 24 is higher than the threshold 20, the process is stopped,
66 IEEE TRANSACTIONS ON ENERGY CONVERSION, VOL. 24, NO. 1, MARCH 2009

Fig. 8. Prototype of the command board.

Fig. 7. Multiple thresholds comparison; y: arbitrary vertical scale.

and the motor turned off. The lower part of the figure shows the
preceding calculations and the profile of the thresholds adopted.
3) Controller of Thresholds: When the speed includes a low-
frequency component associated with a load irregularity, it is
advisable to adapt the series of thresholds S(j). The procedure
can be made automatically due to a threshold controller that
analyzes the speed variations, and decides consequently, to load
a set of precalculated N thresholds rather than another.
In the case of rolling shutters, we adopted M = 4 sets of
N = 18 thresholds. Every 360 ms, the series of N EnvMini is
analyzed and the set of N thresholds is changed if necessary.
The efficiency of the slowing-down detection algorithm
strongly depends on the threshold values. These are determined
on the basis of a statistical study of about a thousand measure- Fig. 9. Speed and voltage of the auxiliary phase V 1 (t), for blocking the shutter
ments, performed on many rolling shutters in operation, with when lowering without cutout.
different motors of the same family, at different temperatures.
The N × M thresholds are calculated with a reciprocal algo-
rithm so that they encompass all the configurations of tested B. Results
rolling shutters, motors, and temperatures, without adjustment In order to validate the slowdown detection method (SDM),
or reprogramming of the command board. thousands of tests have been achieved on various configurations
of loads, motors, and temperatures. Each time, the stoppage
successfully happens, approximately 40–90 ms after the begin-
IV. EXPERIMENTAL VALIDATION ON A ROLLING SHUTTER ning of the slowdown. When the shutter reaches the buffer stop,
the reaction torque on the motor strongly increases. Due to the
A. Equipment celerity of the SDM, this torque is kept lower than its nomi-
The previous study is implemented in practice to detect the nal value (about 0.1 N·m). Without SDM, it would reach the
arrival of a rolling shutter on the buffer stop, without the sensor. maximum (0.2 N·m). Experimentally, this result is satisfactory.
The experimental bench is made up of an SPIM motor supplied The shutter is clamped on the buffer stop without mechanical
by the ac power with an inversion device for the rotational direc- distortion or audible grating.
tion (see Fig. 1). The drive board shown in Fig. 8 is made around A test without SDM is given Fig. 9 and Table III. It allows
an 8-bit microcontroller Microchip PIC16F876. It decodes and to study the whole slowing-down process and measure the ef-
interprets the user orders sent by the radio. Two circuits convert ficiency of an eventual SDM. The speed, the phase difference
and filter the voltages V1 and V2 to 0–5 V compatible levels. arg (V1 /V2 ), and the amplitude of V1 are recorded at the time of
They are connected to the analog-to-digital (ADC) inputs and a forced stoppage. The rotor speed is measured by an incremen-
the two programmable capture inputs of the microcontroller. tal encoder giving one pulse per revolution. The slowing down
This gives access to the phase difference between V1 and V2 and starts a little after t1 = 25.1 s. The last speed measured N2 is at
their amplitudes. t2 = 25.4 s–beyond this, accuracy is no longer ensured.

Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
FERREYRE et al.: SENSORLESS SLOWDOWN DETECTION METHOD
TABLE III
CHARACTERISTIC POINTS OF FIG. 9
Fig. 9 and Table III clearly illustrate that the phase difference
and the norm of V1 have the same characteristics as that of the
speed, with a measurement delay of 10–30 ms. The theoretical
results of Section II are confirmed. On the recording, we can see
that SDM, with the most sensitive of the four thresholds series,
would stop the motor in tS = 25 290 ms, e.g., 60 ms after the
beginning of the slowing down.
When the temperature increases, the curves arg (V1 /V2 ) and
|V1 /V2 | are slightly shifted down, but as the algorithm makes
comparisons between values, this turns out to have no effect on
the criteria.
V. CONCLUSION
This paper shows a system for assessing the speed and de-
tecting the slowing down applied to the single-phase induction
motor with a running capacitor. The appraisal of the speed re-
poses on the measurement of the voltages. The disposition is
researched in numerous industrial scenarios to save on the speed
sensor. The study starts by an analytic modeling of the evolution
in the voltages and currents with the speed. It then appears that
the speed can be estimated on the basis of voltages and currents
of the stator or the capacitor. The value used is determined from
the parameters of the motor.
The implementation is not simple, since the values measured
are disturbed by noise. This paper suggests an original and sys-
tematic method for reducing the impact of noise. It shows an
algorithm that responds straight away to an excessive slowing
down while remaining resistant to the electrical parasites and the
irregularities of speed imposed by the load. Generally speaking,
the thresholds and the background history of the N dimensions
are adjusted according to the load profile, which makes the algo-
rithm suitably adaptable to any type of application. The method
is applied in this case to a rolling shutter motor, the objective
being an end-of-travel stop. The objective has been reached
with an assembly using a microcontroller, and the measurement
results are shown with related comments.
REFERENCES
[1] J. P. Borrelli and R. Burkhart, “New phase-sensitive technology for capac-
itor start motor simplifies application,” IEEE Trans. Ind. Appl., vol. 34,
no. 2, pp. 253–257, Mar./Apr. 1998.
[2] M. B. de Rossiter Corrêa, C. B. Jacobina, E. R. C. da Silva, and
A. M. N. Lima, “Vector control strategies for single-phase induction mo-
tors drive systems,” IEEE Trans. Ind. Electron., vol. 51, no. 5, pp. 1073–
1080, Oct. 2004.
[3] A. Payman and S. Vaez-Sadeh, “DSP based speed estimation of single
phase induction motors,” in Proc. 2004 IEEE 35th Annu. Power Electron.
Spec. Conf. (PESC 2004), vol. 2, pp. 1335–1340.
[4] S. Vaez-Sadeh and S. H. Reicy, “Sensorless vector control of single-phase
induction motor drives,” in Proc. 8th IEEE Int. Conf. Electr. Mach. Syst.
(ICEMS 2005), vol. 3, pp. 1838–1842.
Authorized licensed use limited to: UNIVERSIDADE DE SAO PAULO. Downloaded on May 21, 2009 at 14:52 from IEEE Xplore. Restrictions apply.
67
[5] P. Vas, Electrical Machines and Drives: A Space Vector Theory Approach.
Oxford, U.K.: Clarendon, 1992, pp. 502–511.
[6] H. Huang, E. F. Fuchs, and J. C. White, “Optimal placement of the run
capacitor in single phase induction motor design,” IEEE Trans. Energy
Convers., vol. 3, no. 3, pp. 647–652, Sep. 1988.
[7] E. F. Fuchs, A. J. Vandenput, J. Höll, and J. C. White, “Design and
analysis of capacitor-start, capacitor-run single-phase induction motors,”
IEEE Trans. Energy Convers., vol. 5, no. 2, pp. 327–336, Jun. 1990.
[8] L. Xu, “Dynamic model of an integrated-cycle controlled single phase
induction machine,” IEEE Trans. Energy Convers., vol. 7, no. 4, pp. 761–
767, Dec. 1992.
[9] D.-H. Jang and J.-S. Won, “Voltage, frequency, and phase-difference angle
control of PWM inverters-fed two-phase induction motors,” IEEE Trans.
Power Electron., vol. 9, no. 4, pp. 377–383, Jul. 1994.
[10] C. Mademlis, I. Kioskeridis, and T. Theodoulidis, “Optimisation of single-
phase motors, Part 1: Maximum energy efficiency control,” IEEE Trans.
Energy Convers., vol. 20, no. 1, pp. 187–195, Mar. 2005.
[11] C. Mademlis, I. Kioskeridis, and T. Theodoulidis, “Optimisation of single-
phase motors, Part 2: Maximum energy efficiency control,” IEEE Trans.
Energy Convers., vol. 20, no. 1, pp. 196–203, Mar. 2005.
[12] D. Nehring, “Microprocessor based motor control system with phase dif-
ference detection,” U.S. Patent 5 808 441, Sep. 15, 1998.
[13] D. Nassr, P. Raude, and L. Plumer, “Control device for stopping the
operation of a single-phase asynchronous motor with a capacitor,” U.S.
Patent 5 760 556, Jun. 2, 1998.
[14] P. Raude and A. Mansouri, “Stop control device for single-phase asyn-
chronous motor with capacitor,” French Patent 2 785 469, Mar. 5, 2000.
[15] M. Jacob and S. Dutertre, “Stop control device for single-phase asyn-
chronous motor with capacitor,” French Patent 2 838 886, Oct. 24, 2003.
[16] T. Stempniakowski, “Control device for an electric motor powered by ac,”
French Patent 2 813 453, Mar. 1, 2002.
Frédéric Ferreyre was born in 1982. He received
the Engineering degree from the Ecole Catholique
d’Arts et Métiers de Lyon (ECAM) Engineering
School, Lyon, France, in 2003. He is currently work-
ing toward the Ph.D. degree at the University Claude
Bernard Lyon 1, Villeurbanne, France.
He is also currently with Profalux, where he is
engaged in driving induction motors.
René Goyet was born in France, in 1946. He re-
ceived the Engineering degree from Ecole Speciale
des Travaux Publics de Paris (ESTP), Paris, France,
in 1969, and the “These d’Etat” (long Ph.D.) from
the University of Paris 6, Paris, in 1981.
He was an Assistant Professor at the University of
Paris 6 for 30 years. He is currently with the Univer-
sity Claude Bernard Lyon 1, Villeurbanne, France. He
has also been a teacher of electronics, electrotechnics,
and automatics. His current research interests include
driving motors and magnetic modeling for electrical
engineering. He has authored or coauthored numerous technical publications.
Guy Clerc (M’00) was born in Libourne, France,
on November 30, 1960. He received the Engineering
and Ph.D. degrees in electrical engineering from the
Ecole Centrale de Lyon, Lyon, France, in 1984 and
1989, respectively.
He is currently a Full Professor, and teaches elec-
trical engineering at the University Claude Bernard
Lyon 1 (UCBL), Villeurbanne, France, where he is
engaged in research on control and diagnosis of in-
duction machines at Laboratoire Ampere.
Thierry Bouscasse was born in 1951. He is the Chairman of Profalux, Thyez,
France.